I have Rcpp and RccpEigen already installed in RStudio. I am able to run an Rcpp code (that didn't use RccpEigen) successfully as well. However the following code which uses both doesn't seem to work.
Here is the code -
library(Rcpp)
library(RcppEigen)
sourceCpp(code = '
#include <Rcpp.h>
#include <RcppEigen.h>
// [[Rcpp::depends(RcppEigen)]]
using namespace Rcpp;
using namespace Eigen;
using namespace RcppEigen;
// [[Rcpp::export]]
List luEigen(MatrixXd M) {
FullPivLU<MatrixXd> luE(M);
return List::create(Named("L_matrix") = luE.matrixLU().triangularView<Upper>());
}')
A <- 0.8 + 0.2 * diag(100)
(luEigen(A))
This code gives a really long error, so here are the key error lines -
/Library/Frameworks/R.framework/Versions/4.1/Resources/library/Rcpp/include/Rcpp/generated/Vector__create.h:71:10: note: in instantiation of function template specialization 'Rcpp::Vector<19, PreserveStorage>::create__dispatch<Rcpp::traits::named_object<Eigen::TriangularView<const Eigen::Matrix<double, -1, -1, 0>, 2>>>' requested here
return create__dispatch( typename traits::integral_constant<bool,
^
file16bbd8305f5c.cpp:11:18: note: in instantiation of function template specialization 'Rcpp::Vector<19, PreserveStorage>::create<Rcpp::traits::named_object<Eigen::TriangularView<const Eigen::Matrix<double, -1, -1, 0>, 2>>>' requested here
return List::create(Named("L_matrix") = luE.matrixLU().triangularView<Upper>());
^
18 warnings and 1 error generated.
make: *** [file16bbd8305f5c.o] Error 1
clang++ -mmacosx-version-min=10.13 -std=gnu++14 -I"/Library/Frameworks/R.framework/Resources/include" -DNDEBUG -I"/Library/Frameworks/R.framework/Versions/4.1/Resources/library/Rcpp/include" -I"/Library/Frameworks/R.framework/Versions/4.1/Resources/library/RcppEigen/include" -I"/private/var/folders/_3/wdql3v5d4vggzffw3xdcr3p80000gn/T/RtmpQioi38/sourceCpp-x86_64-apple-darwin17.0-1.0.7" -I/usr/local/include -fPIC -Wall -g -O2 -c file16bbd8305f5c.cpp -o file16bbd8305f5c.o
Given that Rcpp and RccpEigen are installed and a different Rccp code does work, what may be causing error in this code?
With a very helpful suggestion from #Dirk, I simplified the decomposition and that did the trick. Still not sure why the more complex construction threw an error, but bottom line is that simplification gets the job done. Here is my modified code that works -
library(Rcpp)
library(RcppEigen)
sourceCpp(code = '
#include <Rcpp.h>
#include <RcppEigen.h>
// [[Rcpp::depends(RcppEigen)]]
using namespace Rcpp;
using namespace Eigen;
using namespace RcppEigen;
// [[Rcpp::export]]
List luEigen(MatrixXd M) { // here I name our function
FullPivLU<MatrixXd> luE(M); // here I perform the decomposition
MatrixXd upper = luE.matrixLU().triangularView<Upper>(); // this creates the upper matrix
MatrixXd lower = luE.matrixLU().triangularView<StrictlyLower>(); // this creates the lower matrix
return List::create(Named("U_matrix") = upper, Named("L_matrix") = lower); // this makes the list of the 2 matrices
}')
A <- 0.8 + 0.2 * diag(100)
(luEigen(A))
You could possibly speed it up further by doing the decomposition only once and calling the upper and lower triangular matrices from it, like so -
FullPivLU<MatrixXd> luE(M); // here I perform the decomposition
MatrixXd decomp = luE.matrixLU();
MatrixXd upper = decomp.triangularView<Upper>(); // this creates the upper matrix
MatrixXd lower = decomp.triangularView<StrictlyLower>(); // this creates the lower matrix
Related
I am trying to run a very simple model with rstan in batch mode and have some errors. I do not have any errors when I try to run the same code not in batch mode.
The community from STAN Forums helped me to understand that the issue is that Rcpp instead of STAN is not working properly on the server of my university. We reached this understanding from a test trying to run a very simple Rcpp code in batch mode and not in batch mode. The Rcpp code is:
### Libraries
library(Rcpp)
Rcpp::sourceCpp(code='
#include <Rcpp.h>
// [[Rcpp::export]]
int fibonacci(const int x) {
if (x == 0) return(0);
if (x == 1) return(1);
return (fibonacci(x - 1)) + fibonacci(x - 2);
}',
verbose = TRUE,
rebuild = TRUE
)
The results are below.
Not in batch mode:
Generated R functions
-------------------------------------------------------
`.sourceCpp_1_DLLInfo` <- dyn.load('/tmp/RtmpDNL6of/sourceCpp-x86_64-pc-linux-gnu-1.0.7/sourcecpp_178b21d435d91/sourceCpp_4.so')
fibonacci <- Rcpp:::sourceCppFunction(function(x) {}, FALSE, `.sourceCpp_1_DLLInfo`, 'sourceCpp_1_fibonacci')
rm(`.sourceCpp_1_DLLInfo`)
Building shared library
--------------------------------------------------------
DIR: /tmp/RtmpDNL6of/sourceCpp-x86_64-pc-linux-gnu-1.0.7/sourcecpp_178b21d435d91
/hpc/apps/R/4.1.2/lib64/R/bin/R CMD SHLIB --preclean -o 'sourceCpp_4.so' 'file178b25628f5ea.cpp'
g++ -std=gnu++14 -I"/hpc/apps/R/4.1.2/lib64/R/include" -DNDEBUG -I"/hpc/home/user/R/x86_64-pc-linux-gnu-library/4.1/Rcpp/include" -I"/tmp/RtmpDNL6of/sourceCpp-x86_64-pc-linux-gnu-1.0.7" -I/hpc/apps/zlib/1.2.8/include -I/hpc/apps/pcre/8.38/include -I/hpc/apps/bzip2/1.0.6/include -I/hpc/apps/curl/7.54.1/include -I/hpc/apps/cairo/1.14.12/include -fpic -g -O2 -c file178b25628f5ea.cpp -o file178b25628f5ea.o
g++ -std=gnu++14 -shared -L/hpc/apps/R/4.1.2/lib64/R/lib -L/hpc/apps/zlib/1.2.8/lib -L/hpc/apps/pcre/8.38/lib -L/hpc/apps/bzip2/1.0.6/lib -L/hpc/apps/curl/7.54.1/lib -L/hpc/apps/cairo/1.14.12/lib -o sourceCpp_4.so file178b25628f5ea.o -L/hpc/apps/R/4.1.2/lib64/R/lib -lR
In batch mode:
Generated R functions
-------------------------------------------------------
`.sourceCpp_1_DLLInfo` <- dyn.load('/localscratch/250948.1.ragatkolab.q/RtmpDlK5Sy/sourceCpp-x86_64-pc-linux-gnu-1.0.7/sourcecpp_142bd68d660f1/sourceCpp_2.so')
fibonacci <- Rcpp:::sourceCppFunction(function(x) {}, FALSE, `.sourceCpp_1_DLLInfo`, 'sourceCpp_1_fibonacci')
rm(`.sourceCpp_1_DLLInfo`)
Building shared library
--------------------------------------------------------
DIR: /localscratch/250948.1.mylab.q/RtmpDlK5Sy/sourceCpp-x86_64-pc-linux-gnu-1.0.7/sourcecpp_142bd68d660f1
/hpc/apps/R/4.1.2/lib64/R/bin/R CMD SHLIB -o 'sourceCpp_2.so' 'file142bd64ebe902.cpp'
g++ -std=gnu++14 -I"/hpc/apps/R/4.1.2/lib64/R/include" -DNDEBUG -I"/hpc/home/user/R/x86_64-pc-linux-gnu-library/4.1/Rcpp/include" -I"/localscratch/250948.1.ragatkolab.q/RtmpDlK5Sy/sourceCpp-x86_64-pc-linux-gnu-1.0.7" -I/hpc/apps/zlib/1.2.8/include -I/hpc/apps/pcre/8.38/include -I/hpc/apps/bzip2/1.0.6/include -I/hpc/apps/curl/7.54.1/include -I/hpc/apps/cairo/1.14.12/include -fpic -g -O2 -c file142bd64ebe902.cpp -o file142bd64ebe902.o
In file included from /hpc/apps/gcc/9.1.0/include/c++/9.1.0/x86_64-pc-linux-gnu/bits/c++config.h:524,
from /hpc/apps/gcc/9.1.0/include/c++/9.1.0/cmath:41,
from /hpc/home/user/R/x86_64-pc-linux-gnu-library/4.1/Rcpp/include/Rcpp/platform/compiler.h:100,
from /hpc/home/user/R/x86_64-pc-linux-gnu-library/4.1/Rcpp/include/Rcpp/r/headers.h:66,
from /hpc/home/user/R/x86_64-pc-linux-gnu-library/4.1/Rcpp/include/RcppCommon.h:30,
from /hpc/home/user/R/x86_64-pc-linux-gnu-library/4.1/Rcpp/include/Rcpp.h:27,
from file142bd64ebe902.cpp:2:
/hpc/apps/gcc/9.1.0/include/c++/9.1.0/x86_64-pc-linux-gnu/bits/os_defines.h:39:10: fatal error: features.h: No such file or directory
39 | #include <features.h>
| ^~~~~~~~~~~~
compilation terminated.
make: *** [file142bd64ebe902.o] Error 1
Error in Rcpp::sourceCpp(code = "\n #include <Rcpp.h>\n\n // [[Rcpp::export]]\n int fibonacci(const int x) {\n if (x == 0) return(0);\n if (x == 1) return(1);\n return (fibonacci(x - 1)) + fibonacci(x - 2);\n }", :
Error 1 occurred building shared library.
WARNING: The tools required to build C++ code for R were not found.
Please install GNU development tools including a C++ compiler.
Execution halted
Then, they asked me to modify my makevars file - which I did using the R-package usethis::edit_r_makevars() - adding the lines:
CXXFLAGS += -I"/hpc/apps/gcc/9.1.0/include/c++/9.1.0/parallel/"
CXX14FLAGS += -I"/hpc/apps/gcc/9.1.0/include/c++/9.1.0/parallel/"
Once I did this modification (I hope I did it correctly), my error message change to:
Generated extern "C" functions
--------------------------------------------------------
#include <Rcpp.h>
#ifdef RCPP_USE_GLOBAL_ROSTREAM
Rcpp::Rostream<true>& Rcpp::Rcout = Rcpp::Rcpp_cout_get();
Rcpp::Rostream<false>& Rcpp::Rcerr = Rcpp::Rcpp_cerr_get();
#endif
// fibonacci
int fibonacci(const int x);
RcppExport SEXP sourceCpp_1_fibonacci(SEXP xSEXP) {
BEGIN_RCPP
Rcpp::RObject rcpp_result_gen;
Rcpp::RNGScope rcpp_rngScope_gen;
Rcpp::traits::input_parameter< const int >::type x(xSEXP);
rcpp_result_gen = Rcpp::wrap(fibonacci(x));
return rcpp_result_gen;
END_RCPP
}
Generated R functions
-------------------------------------------------------
`.sourceCpp_1_DLLInfo` <- dyn.load('/localscratch/251578.1.mylablab.q/RtmpCHyEAd/sourceCpp-x86_64-pc-linux-gnu-1.0.7/sourcecpp_1e50364a03315/sourceCpp_2.so')
fibonacci <- Rcpp:::sourceCppFunction(function(x) {}, FALSE, `.sourceCpp_1_DLLInfo`, 'sourceCpp_1_fibonacci')
rm(`.sourceCpp_1_DLLInfo`)
Building shared library
--------------------------------------------------------
DIR: /localscratch/251578.1.mylablab.q/RtmpCHyEAd/sourceCpp-x86_64-pc-linux-gnu-1.0.7/sourcecpp_1e50364a03315
/hpc/apps/R/4.1.2/lib64/R/bin/R CMD SHLIB -o 'sourceCpp_2.so' 'file1e50325e7686.cpp'
g++ -std=gnu++14 -I"/hpc/apps/R/4.1.2/lib64/R/include" -DNDEBUG -I"/hpc/home/user/R/x86_64-pc-linux-gnu-library/4.1/Rcpp/include" -I"/localscratch/251578.1.mylablab.q/RtmpCHyEAd/sourceCpp-x86_64-pc-linux-gnu-1.0.7" -I/hpc/apps/zlib/1.2.8/include -I/hpc/apps/pcre/8.38/include -I/hpc/apps/bzip2/1.0.6/include -I/hpc/apps/curl/7.54.1/include -I/hpc/apps/cairo/1.14.12/include -fpic -g -O2 -I"/hpc/apps/gcc/9.1.0/include/c++/9.1.0/parallel/" -c file1e50325e7686.cpp -o file1e50325e7686.o
In file included from /hpc/apps/gcc/11.1.0/include/c++/11.1.0/x86_64-pc-linux-gnu/bits/c++config.h:571,
from /hpc/apps/gcc/11.1.0/include/c++/11.1.0/cmath:41,
from /hpc/home/user/R/x86_64-pc-linux-gnu-library/4.1/Rcpp/include/Rcpp/platform/compiler.h:100,
from /hpc/home/user/R/x86_64-pc-linux-gnu-library/4.1/Rcpp/include/Rcpp/r/headers.h:66,
from /hpc/home/user/R/x86_64-pc-linux-gnu-library/4.1/Rcpp/include/RcppCommon.h:30,
from /hpc/home/user/R/x86_64-pc-linux-gnu-library/4.1/Rcpp/include/Rcpp.h:27,
from file1e50325e7686.cpp:2:
/hpc/apps/gcc/11.1.0/include/c++/11.1.0/x86_64-pc-linux-gnu/bits/os_defines.h:44:19: error: missing binary operator before token "("
44 | #if __GLIBC_PREREQ(2,15) && defined(_GNU_SOURCE)
| ^
/hpc/apps/gcc/11.1.0/include/c++/11.1.0/x86_64-pc-linux-gnu/bits/os_defines.h:52:19: error: missing binary operator before token "("
52 | #if __GLIBC_PREREQ(2, 27)
| ^
In file included from /hpc/home/user/R/x86_64-pc-linux-gnu-library/4.1/Rcpp/include/Rcpp/platform/compiler.h:100,
from /hpc/home/user/R/x86_64-pc-linux-gnu-library/4.1/Rcpp/include/Rcpp/r/headers.h:66,
from /hpc/home/user/R/x86_64-pc-linux-gnu-library/4.1/Rcpp/include/RcppCommon.h:30,
from /hpc/home/user/R/x86_64-pc-linux-gnu-library/4.1/Rcpp/include/Rcpp.h:27,
from file1e50325e7686.cpp:2:
/hpc/apps/gcc/11.1.0/include/c++/11.1.0/cmath:45:15: fatal error: math.h: No such file or directory
45 | #include_next <math.h>
| ^~~~~~~~
compilation terminated.
make: *** [file1e50325e7686.o] Error 1
Error in Rcpp::sourceCpp(code = "\n #include <Rcpp.h>\n\n // [[Rcpp::export]]\n int fibonacci(const int x) {\n if (x == 0) return(0);\n if (x == 1) return(1);\n return (fibonacci(x - 1)) + fibonacci(x - 2);\n }", :
Error 1 occurred building shared library.
WARNING: The tools required to build C++ code for R were not found.
Please install GNU development tools including a C++ compiler.
Execution halted
My knowledge in Linux is limited and I am afraid that my question is out of scope for STAN Forums. Any thoughts how to proceed?
I fear your problem is still local. On the computing you use, someone may have set up R in way that makes it differ between interactive R use, and what you call batch use.
But that is non-standard, and at your end. In general, it just works. Using a minimally modified version of your program (which I still find confusing: it loads doParallel, registers cores but ... triggers to parallel code -- anyway) it works just fine:
Using Rscript
edd#rob:~/git/stackoverflow/70470842(master)$ Rscript answer.R
Loading required package: foreach
Loading required package: iterators
Loading required package: parallel
Fib(10) is 55
edd#rob:~/git/stackoverflow/70470842(master)$
Using R in batch mode
edd#rob:~/git/stackoverflow/70470842(master)$ R -s -f answer.R
Loading required package: foreach
Loading required package: iterators
Loading required package: parallel
Fib(10) is 55
edd#rob:~/git/stackoverflow/70470842(master)$
Using our littler frontend
edd#rob:~/git/stackoverflow/70470842(master)$ r answer.R
Loading required package: foreach
Loading required package: iterators
Loading required package: utils
Loading required package: parallel
Fib(10) is 55
edd#rob:~/git/stackoverflow/70470842(master)$
Modified source code
library(doParallel) # Library
ncores <- 5 # Simulation parameters
n_sim <- 5
registerDoParallel(ncores)
Rcpp::sourceCpp(code='
#include <Rcpp.h>
// [[Rcpp::export]]
int fibonacci(const int x) {
if (x == 0) return(0);
if (x == 1) return(1);
return (fibonacci(x - 1)) + fibonacci(x - 2);
}'
)
cat("Fib(10) is ", fibonacci(10), "\n")
For future reference, I interacted with IT team from my institution after feedback received here and STAN Forums, and they finally found a solution. In their words:
"Compute nodes aren't normally used for development. So certain development packages are not installed by default. We installed "glibc-devel" and "glibc-headers" on the compute nodes.
So, it seems that when run interactively the job was running on the submit node (which had these libraries installed) and when run from the scheduler it was run on the compute nodes (which did not have the library installed). The reason for the discrepancy being that the libraries in question have traditionally been only used during development and compilation. This use posed a unique use case."
Rcpp is powerful and has worked great in most all cases, but I cannot figure out how to wrap a C-function that returns a user-defined structure into an R package.
DESCRIPTION
Package: myPackage
Type: Package
Title: What the Package Does (Title Case)
Version: 0.1.0
Depends: R (>= 4.0.0)
License: GPL-3
Encoding: UTF-8
LazyData: true
RoxygenNote: 7.1.0
Imports:
Rcpp
Suggests:
knitr,
rmarkdown,
testthat
VignetteBuilder: knitr
LinkingTo:
Rcpp
NAMESPACE
# Generated by roxygen2: do not edit by hand
importFrom(Rcpp,evalCpp)
export(read_header)
useDynLib(myPackage, .registration = TRUE)
Header File ("inst/include/myPackage_types.h")
#include <Rcpp.h>
namespace Rcpp {
typedef struct {
int my_data;
} MY_HEADER_INFO;
template <> SEXP wrap(const MY_HEADER_INFO& x) {
std::vector<std::string> names;
std::vector<SEXP> elements(1);
// do something with the elements and names
names.push_back("my_data");
elements[0] = Rcpp::wrap( x.my_data );
return 0;
};
}
C-code
/*
read_header.c
*/
#include <stdio.h>
#include <stdlib.h>
#include <Rcpp.h>
#include "inst/include/myPackage_types.h"
namespace Rcpp {
// [[Rcpp::export]]
Rcpp::MY_HEADER_INFO read_header() {
Rcpp::MY_HEADER_INFO *header;
header = (Rcpp::MY_HEADER_INFO*)malloc(sizeof(Rcpp::MY_HEADER_INFO));
memset(header, 0, sizeof(Rcpp::MY_HEADER_INFO));
return *header;
}
}
When I source (i.e., compile) the C-code, I get the following:
Error in dyn.load("/private/var/folders/gl/jvj9b0xn34lgq6_h9370p8q80000gn/T/RtmpYRaBGW/sourceCpp-x86_64-apple-darwin17.0-1.0.4.6/sourcecpp_25a13fe3d7da/sourceCpp_49.so") :
unable to load shared object ...
When I try to build the package (CMD + Shift + B), I get:
clang++ -mmacosx-version-min=10.13 -std=gnu++11 -I"/Library/Frameworks/R.framework/Resources/include" -DNDEBUG -I'/Library/Frameworks/R.framework/Versions/4.0/Resources/library/Rcpp/include' -I/usr/local/include -fPIC -Wall -g -O2 -UNDEBUG -Wall -pedantic -g -O0 -fdiagnostics-color=always -c RcppExports.cpp -o RcppExports.o
RcppExports.cpp:9:7: error: no type named 'MY_HEADER_INFO' in namespace 'Rcpp'
Rcpp::MY_HEADER_INFO read_header();
~~~~~~^
This seems so simple, but I can't find a similar vignetter or example and none of the variations I've tried seem to work. What is the cause of the dynamical loading error for SourceCpp? Why can't the compiler find the code in the header file?
Thanks!
I got this to work and have posted my solution to GitHub: git#github.com:markrbower/myPackage.git
The key parts are:
inst/include/myPackage_types.h
#include <RcppCommon.h>
namespace Rcpp {
typedef struct {
int my_data;
} MY_HEADER_INFO;
template <> SEXP wrap(const MY_HEADER_INFO& x);
template<> MY_HEADER_INFO* as(SEXP x);
}
read_header.cpp
/*
read_header.c
*/
#include <stdio.h>
#include <stdlib.h>
#include "../inst/include/myPackage_types.h"
#include <RcppCommon.h>
namespace Rcpp {
template <> SEXP wrap(const MY_HEADER_INFO& x);
}
#include <Rcpp.h>
namespace Rcpp {
template <> SEXP wrap(const MY_HEADER_INFO& x) {
std::vector<std::string> names;
std::vector<SEXP> elements(1);
// do something with the elements and names
names.push_back("my_data");
elements[0] = wrap( x.my_data );
Rcpp::List result(elements.size());
for (size_t i = 0; i < elements.size(); ++i) {
result[i] = elements[i];
}
result.attr("names") = Rcpp::wrap(names);
// result can be return to R as a list
return( result );
};
}
//' #importFrom Rcpp evalCpp
//' #useDynLib myPackage
//' #export
// [[Rcpp::export]]
Rcpp::MY_HEADER_INFO read_header() {
Rcpp::MY_HEADER_INFO *header = NULL;
printf( "%ld\n", sizeof(Rcpp::MY_HEADER_INFO) );
header = (Rcpp::MY_HEADER_INFO*)malloc(sizeof(Rcpp::MY_HEADER_INFO));
memset(header, 0, sizeof(Rcpp::MY_HEADER_INFO));
header->my_data = 10;
return *header;
}
There were two problems. First, I had template definitions under the wrong Rcpp header (put the initial calls after RcppCommon.h and the detailed calls after Rcpp.h). The tutorials and examples warned me not to do that, but I did it, anyway. Second, I found that if you "source" the code and then "load" the library, the sourced code will obscure the library code and you will get a "null pointer" error. Running "devtools::check()" showed me that along with noting that the fix is to "rm" the sourced function.
Let me also add that there are two Roxygen comments that I needed to add to my .cpp file to get the appropriate commands to appear in my NAMESPACE file:
//' #importFrom Rcpp evalCpp
//' #useDynLib myPackage
I found an old post where Dirk suggested using the Rcpp.package.skeleton function to build a "baby" project and then slowly add things until you can do what you want. That is so much better than the approach I have been using: Start with a complex C program and try to shoehorn my code into Rcpp.
Closed. This question is not reproducible or was caused by typos. It is not currently accepting answers.
This question was caused by a typo or a problem that can no longer be reproduced. While similar questions may be on-topic here, this one was resolved in a way less likely to help future readers.
Closed 4 years ago.
Improve this question
I have some functions written in Rcpp and RcppArmadillo like this
example.cpp:
#include <RcppArmadillo.h>
// [[Rcpp::depends(RcppArmadillo)]]
#include <iostream>
#include <math.h>
using namespace Rcpp;
// using namespace RcppArmadillo;
using namespace arma;
using namespace std;
// [[Rcpp::export]]
double inner1(NumericVector x, NumericVector y) {
int K = x.length() ;
double ip = 0 ;
for (int k = 0 ; k < K ; k++) {
ip += x(k) * y(k) ;
}
return(ip) ;
}
// [[Rcpp::export]]
mat multiply2(mat A, mat B) {
return A * B;
}
Then I use Rcpp::sourceCpp('example.cpp') and It works well under Ubuntu.
(I have run library(Rcpp) and library(RcppArmadillo) before )
However, When I move to Windows Platform, the RStudio Throw an error said:
> Rcpp::sourceCpp('R:/example.cpp')
Error in inDL(x, as.logical(local), as.logical(now), ...) :
unable to load shared object 'C:/Users/[Username]/AppData/Local/Temp/RtmpG6H80X/sourceCpp-x86_64-w64-mingw32-0.12.19/sourcecpp_40b04b2c2bcf/sourceCpp_4.dll':
LoadLibrary failure: The specified procedure could not be found.
I found the key problem is in the matrix multiplication. Since I try to delete the second function multiply2. Then the rest of the code can be complied successfully under Windows.
example2.cpp
#include <RcppArmadillo.h>
// [[Rcpp::depends(RcppArmadillo)]]
#include <iostream>
#include <math.h>
using namespace Rcpp;
// using namespace RcppArmadillo;
using namespace arma;
using namespace std;
// [[Rcpp::export]]
double inner1(NumericVector x, NumericVector y) {
int K = x.length() ;
double ip = 0 ;
for (int k = 0 ; k < K ; k++) {
ip += x(k) * y(k) ;
}
return(ip) ;
}
I have tried some other codes, found that this error occurs when matrix multiplication * was used in code.
So, Why matrix multiplication in RcppArmadillo fails under Windows Platform?
After the struggled with the compiler for a long time, I found the key point is the BLAS library was not correctly set in my windows system environment path.
In short, the solution is :
Download precompiled binary packages of OpenBLS at Here (DO NOT try to compile the newest version under windows, I've waste a lot time on it)
Exactor OpenBLAS-v0.2.19-Win64-int32.zip to some where like C:\LIBS\OpenBLAS-v0.2.15-Win64-int32
Add C:\LIBS\OpenBLAS-v0.2.15-Win64-int32\bin to your PATH
[Optional] Create a new environment variable called BLAS_LIBS, which value is C:\LIBS\OpenBLAS-v0.2.15-Win64-int32\bin
Restart RStudio, and the problem was solved.
I find this solution by install RcppArmadillo from source by install.packages("RcppArmadillo", type = "source"), this time the RStudio throw the same error during the compiling, so the install fails.
However, if I just use install.packages("RcppArmadillo"), the RStudio will install binary version of RcppArmadillo, so I didn't get any feedback about the absence of BLAS.
I was trying to run this peace of code in R (credit to the author):
require(Rcpp)
require(RcppArmadillo)
require(inline)
cosineRcpp <- cxxfunction(
signature(Xs = "matrix"),
plugin = c("RcppArmadillo"),
body='
Rcpp::NumericMatrix Xr(Xs); // creates Rcpp matrix from SEXP
int n = Xr.nrow(), k = Xr.ncol();
arma::mat X(Xr.begin(), n, k, false); // reuses memory and avoids extra copy
arma::mat Y = arma::trans(X) * X; // matrix product
arma::mat res = (1 - Y / (arma::sqrt(arma::diagvec(Y)) * arma::trans(arma::sqrt(arma::diagvec(Y)))));
return Rcpp::wrap(res);
')
And got, after few fixes, the following error:
Error in compileCode(f, code, language = language, verbose = verbose) :
Compilation ERROR, function(s)/method(s) not created!
clang: error: no such file or directory: '/usr/local/lib/libfontconfig.a'
clang: error: no such file or directory: '/usr/local/lib/libreadline.a'
make: *** [file5a681e35ebe1.so] Error 1
In addition: Warning message:
running command '/Library/Frameworks/R.framework/Resources/bin/R CMD SHLIB file5a681e35ebe1.cpp 2> file5a681e35ebe1.cpp.err.txt' had status 1
I used to use Rcpp a lot in the past. But between now and then my computer has been reformatted and all the installation re-done using homebrew.
I installed cairo with brew: brew install cairo
the libreadline.a error was solved with:
brew link --force readline
But the same did not work for libfontconfig.a since was already linked:
brew link --force fontconfig
Warning: Already linked: /usr/local/Cellar/fontconfig/2.11.1
To relink: brew unlink fontconfig && brew link fontconfig
I would have assumed that fontconfig is within cairo. In fact, when I type
brew install fontconfig
Warning: fontconfig-2.11.1 already installed
But the truth is that there is no libfontconfig.a at /usr/local/lib/:
ls /usr/local/lib/libfont*
/usr/local/lib/libfontconfig.1.dylib
/usr/local/lib/libfontconfig.dylib
Using the very questionable approach of going here and download it, the code runs, but still gives a the corresponding warning, since the file corresponds to a different os.x architecture (I did not found one for 10.9):
+ . + ld: warning: ignoring file /usr/local/lib/libfontconfig.a, missing required architecture x86_64 in file /usr/local/lib/libfontconfig.a (2 slices)
So at this stage I am a little lost.
How do I install libfontconfig.a or find the 10.9 version?
In case is of any use, I have Xcode installed, I am on a Mac 10.9.5,
and based on this very nice and detailed answer my ~/.R/Makevars file looks like:
CC=clang
CXX=clang++
FLIBS=-L/usr/local/bin/
Your system setup is broken. Neither R nor Rcpp have anything to do with clang (unless you chose clang as your system compiler) or fontconfig.
So start simpler:
R> library(Rcpp)
R> evalCpp("2 + 2")
[1] 4
R>
This just showed that my system has a working compiler R (and Rcpp) can talk to. We can it more explicit:
R> evalCpp("2 + 2", verbose=TRUE)
Generated code for function definition:
--------------------------------------------------------
#include <Rcpp.h>
using namespace Rcpp;
// [[Rcpp::export]]
SEXP get_value(){ return wrap( 2 + 2 ) ; }
No rebuild required (use rebuild = TRUE to force a rebuild)
[1] 4
R>
and R is clever enough not to rebuild. We can then force a build
R> evalCpp("2 + 2", verbose=TRUE, rebuild=TRUE)
Generated code for function definition:
--------------------------------------------------------
#include <Rcpp.h>
using namespace Rcpp;
// [[Rcpp::export]]
SEXP get_value(){ return wrap( 2 + 2 ) ; }
Generated extern "C" functions
--------------------------------------------------------
#include <Rcpp.h>
// get_value
SEXP get_value();
RcppExport SEXP sourceCpp_0_get_value() {
BEGIN_RCPP
Rcpp::RObject __result;
Rcpp::RNGScope __rngScope;
__result = Rcpp::wrap(get_value());
return __result;
END_RCPP
}
Generated R functions
-------------------------------------------------------
`.sourceCpp_0_DLLInfo` <- dyn.load('/tmp/Rtmpeuaiu4/sourcecpp_6a7c7c8295fc/sourceCpp_2.so')
get_value <- Rcpp:::sourceCppFunction(function() {}, FALSE, `.sourceCpp_0_DLLInfo`, 'sourceCpp_0_get_value')
rm(`.sourceCpp_0_DLLInfo`)
Building shared library
--------------------------------------------------------
DIR: /tmp/Rtmpeuaiu4/sourcecpp_6a7c7c8295fc
/usr/lib/R/bin/R CMD SHLIB -o 'sourceCpp_2.so' --preclean 'file6a7c6d1fc2d6.cpp'
ccache g++ -I/usr/share/R/include -DNDEBUG -I"/usr/local/lib/R/site-library/Rcpp/include" -I"/tmp/Rtmpeuaiu4" -fpic -g -O3 -Wall -pipe -Wno-unused -pedantic -c file6a7c6d1fc2d6.cpp -o file6a7c6d1fc2d6.o
g++ -shared -L/usr/lib/R/lib -Wl,-Bsymbolic-functions -Wl,-z,relro -o sourceCpp_2.so file6a7c6d1fc2d6.o -L/usr/lib/R/lib -lR
[1] 4
R>
and on that you see system details on my side (Linux, also using ccache) that will be different for you.
After that, try (Rcpp)Armadillo one-liners and so on.
I got this Rcpp implementation to the rmvnorm function of the mvtnorm package, and I was wondering what I'd need to add in order for it to use openmp so it can take advantage of multiple cores.
I though this ought to do it:
library(Rcpp)
library(RcppArmadillo)
library(inline)
settings <- getPlugin("RcppArmadillo")
settings$env$PKG_CXXFLAGS <- paste('-fopenmp', settings$env$PKG_CXXFLAGS)
settings$env$PKG_LIBS <- paste('-fopenmp -lgomp', settings$env$PKG_LIBS)
code <- '
#include <omp.h>
using namespace Rcpp;
int cores = 1;
cores = as<int>(cores_);
omp_set_num_threads(cores);
int n = as<int>(n_);
arma::vec mu = as<arma::vec>(mu_);
arma::mat sigma = as<arma::mat>(sigma_);
int ncols = sigma.n_cols;
#pragma omp parallel for schedule(static)
arma::mat Y = arma::randn(n, ncols);
return wrap(arma::repmat(mu, 1, n).t() + Y * arma::chol(sigma));
'
rmvnorm.rcpp <- cxxfunction(signature(n_="integer", mu_="numeric", sigma_="matrix", cores_="integer"), body=code, plugin="RcppArmadillo", settings=settings, verbose=TRUE)
But apparently I'm wrong as I'm getting this compilation error message:
Compilation argument:
/software/free/Linux/redhat_5_x86_64/pkgs/r_3.0.2/lib64/R/bin/R CMD SHLIB file50825babe43a.cpp 2> file50825babe43a.cpp.err.txt
/software/free/Linux/redhat_5_x86_64/pkgs/gcc_4.5.3/bin/g++ -I/software/free/Linux/redhat_5_x86_64/pkgs/r_3.0.2/lib64/R/include -DNDEBUG -I/usr/local/include -I"/software/free/Linux/redhat_5_x86_64/pkgs/r_3.0.2/lib64/R/library/RcppArmadillo/include" -I"/software/free/Linux/redhat_5_x86_64/pkgs/r_3.0.2/lib64/R/library/Rcpp/include" -fopenmp -fpic -g -O2 -c file50825babe43a.cpp -o file50825babe43a.o
In file included from file50825babe43a.cpp:31:0:
/software/free/Linux/redhat_5_x86_64/pkgs/gcc_4.5.3/lib/gcc/x86_64-unknown-linux-gnu/4.5.3/include/omp.h: In function 'SEXPREC* file50825babe43a(SEXPREC*, SEXPREC*, SEXPREC*, SEXPREC*)':
/software/free/Linux/redhat_5_x86_64/pkgs/gcc_4.5.3/lib/gcc/x86_64-unknown-linux-gnu/4.5.3/include/omp.h:56:8: error: expected unqualified-id before string constant
file50825babe43a.cpp:35:26: error: 'omp_set_num_threads' was not declared in this scope
file50825babe43a.cpp:41:1: error: for statement expected before 'arma'
make: *** [file50825babe43a.o] Error 1
ERROR(s) during compilation: source code errors or compiler configuration errors!
.
.
.
Error in compileCode(f, code, language = language, verbose = verbose) :
Compilation ERROR, function(s)/method(s) not created! In file included from file50825babe43a.cpp:31:0:
/software/free/Linux/redhat_5_x86_64/pkgs/gcc_4.5.3/lib/gcc/x86_64-unknown-linux-gnu/4.5.3/include/omp.h: In function 'SEXPREC* file50825babe43a(SEXPREC*, SEXPREC*, SEXPREC*, SEXPREC*)':
/software/free/Linux/redhat_5_x86_64/pkgs/gcc_4.5.3/lib/gcc/x86_64-unknown-linux-gnu/4.5.3/include/omp.h:56:8: error: expected unqualified-id before string constant
file50825babe43a.cpp:35:26: error: 'omp_set_num_threads' was not declared in this scope
file50825babe43a.cpp:41:1: error: for statement expected before 'arma'
make: *** [file50825babe43a.o] Error 1
I may be missing something trivial but I don't know what it is.
This has been covered before:
There are two posts at the Rcpp Gallery about OpenMP and Rcpp
There is also an (older, may need an update) example directory OpenMP in the sources which also gets copied into your installation
And Rcpp Attributes defines a plugin for OpenMP